基于机器视觉的篮球投篮打手动作识别方法  

Method of Recognizing Basketball Shooter Action Based on Machine Vision

作　　者：童金茂[1] TONG Jin-mao(Ministry of Sports,College of General Education,Fujian Chuanzheng Communications College,Fuzhou 350007,China)

机构地区：[1]福建船政交通职业学院通识教育学院体育部,福建福州350007

出　　处：《兰州文理学院学报（自然科学版）》2022年第2期86-90,共5页Journal of Lanzhou University of Arts and Science(Natural Sciences)

基　　金：全国职业教育科学研究规划课题(2020QZJ465)。

摘　　要：传统动作识别方法无法实时在复杂环境背景下进行精准的投篮打手动作识别,因此设计了基于机器视觉的篮球投篮打手动作识别方法.采用八叉树量化方法对彩色篮球投篮打手图像进行量化处理,借助矢量中值删除图像中的噪声,实现图像预处理.将经过预处理的图像作为底层特征,对其进行深度特征提取,得到中层特征描述子.将中层特征设定为局部动作表征算子,通过soft-VLAD算法对局部算子进行高时空特征表述,进而完成篮球投篮打手动作识别.实验结果表明,基于机器视觉的篮球投篮打手动作识别方法能够准确识别篮球投篮打手动作.The traditional action recognition methods can not accurately recognize the action of basketball shooters in real time under the background of complex environment.Therefore,a basketball shooter action recognition method based on machine vision is designed.The octree quantization method is used to quantify the color basketball shooter image,and the image preprocessing is realized by deleting the noise in the image with the help of vector median.The preprocessed image is used as the bottom feature,and the depth feature is extracted to obtain the middle feature descriptor.The middle-level feature is set as the local action representation operator,and the high temporal and spatial feature representation of the local operator is carried out through the soft-Vlad algorithm,so as to complete the action recognition of basketball shooters.The experimental results show that the basketball shooter action recognition method based on machine vision can accurately recognize the basketball shooter action.

关 键 词：机器视觉 篮球投篮 打手动作识别 

分 类 号：G841[文化科学—体育训练] TN911.22[文化科学—体育学]